Bi-Modal Self-Sensing Proprioception and Contact Detection in Dielectric Elastomer Soft Robots

The development of accurate sensing strategies for soft robotic systems represents an ongoing challenge. In general, it is difficult to integrate a sensory system that preserves the robot's compliance while simultaneously providing accurate measurements. Dielectric elastomer (DE) transducers offer a potential solution to this problem owing to their ability to function as soft actuators and sensors at the same time, thus enabling the realization of self-sensing soft robotic systems. In this work, we propose a bi-modal sensory scheme for a DE-based soft robot, which allows simultaneous estimation of the robot's own configuration and detection of external contacts without relying on any external sensors. Self-sensing proprioception and contact detection are performed at the same time while actuating the robot, based on online processing of electrical signals measured directly at the actuators' level, i.e., DEs' voltage and current. Both estimation tasks are achieved in real time by means of artificial neural networks. After describing the system and the bi-modal self-sensing architecture, the approach is validated experimentally on a DE soft robot.